This study employs advanced computer vision techniques to quantify social polarization in Israel by analyzing interpersonal distances among individuals of diverse religious backgrounds in urban public spaces. Given Israel’s pronounced levels of polarization and the significant social implications of the Israeli-Palestinian conflict, assessing social divisions through proxemics offers critical insights. Our analysis utilizes (number of frames) extracted from video recordings of pedestrian activity throughout the day in the cities of Jerusalem, Haifa, Nazareth, and Tel Aviv.
To achieve this, we utilized YOLOv9, a state-of-the-art object detection model renowned for its speed and accuracy in real-time applications. After evaluating multiple datasets, YOLOv9 was identified as the most effective model, balancing both performance and runtime. Additionally, an extensive comparison of various depth estimation models demonstrated that DepthAnything, a versatile model optimized for general-purpose depth estimation in outdoor environments, exhibited the highest accuracy following fine-tuning for outdoor detection.
This methodological approach builds on prior research linking physical distance to social perception, enabling our study to provide empirical insights into the dynamics of social polarization in Israel.